Guest Editorial: Learning-Based Edge Computing Services

Min Chen, Haiyang Wang, Kai Hwang, Giancarlo Fortino, Jeungeun Song, Limei Peng, Joze Guna
2021 IEEE Network  
T he recent revival of artificial intelligence (AI) is revolutionizing almost every branch of science and technology. Given the ubiquitous smart mobile terminals and Internet of Things (IoT) devices, it is expected that a majority of intelligent applications will be deployed at the edge networks. Therefore, providing intelligent and sustainable edge computing services will be a hot topic in IoT and 5G services as sustainability of edge systems becomes a necessity with the rapid constant growth
more » ... f edge devices/sensors. Edge learning is needed to support edge computing services. However, it faces great challenges when considering the latency-sensitive requirements, network bandwidth limitation, computing power limitation, as well as limited data at each edge device. AI-based technologies, such as statistical learning, feedforward neural networks, and deep recurrent neural networks, among others, are expected to construct the intelligent edge for improving Quality of Services (QoS) and Quality of Experience (QoE). These learning-based methods are implemented for sophisticated decision making, network management, resource optimization, and in-depth knowledge discovery in complex environments. The intelligence built by using edge learning techniques can help promote better decision making and contribute to building greener and more sustainable systems. To address several major issues regarding the sustainability of edge computing services, this special issue highlights edge learning techniques to provide intelligent and greener edge computing services. Learning-based approaches are required to obtain more clear and in-depth knowledge of the behavior of edge networks. Submissions are expected to address how to build greener and more sustainable edge systems through learning-based approaches. The scope of this Special Issue (SI) is to present and highlight the advances and the latest intelligent technologies, implementations and applications in the fi eld of learning-based edge computing services, so as to build greener and more sustainable edge systems. The articles in this special issue are classifi ed into the following categories: • Innovative architectures, frameworks, and models for learning-based edge computing. • Theory, standards, protocols, and strategies for learning-based edge computing services in IoT. • Machine learning, AI and other innovative optimization approaches for learning-based edge computing services. • Intelligent and interactive IoT services and applications assisted by machine learning and edge computing. • Intelligent decision-making systems for learning-based edge computing services. • Smart task caching at edges by joint optimization of computation, caching and communication. • Security and privacy assisted by learning-based edge computing in IoT. • Learning-based testbeds, simulations, experiments and evaluation for edge computing. In the fi rst article, "Edge-Learning-Enabled Realistic Touch and Stable Communication for Remote Haptic Display", the authors propose a novel remote haptic display system enabled by an edge computing platform, and learning-based methods are employed for realistic haptic feedback and stable haptic communication. More specifically, the control algorithm combining PID neural network with decoupling is applied for fast and accurate generation of the referenced magnetic fi eld at the haptic device, and a supervised BiLSTM network with user operations input is constructed for the prediction of missing haptic data in remote transmission. The developed system provides a practical platform for real-time remote haptic data acquisition and display, and experimental results demonstrate the eff ective sustainability of remote haptic interaction service. In the second article, "Toward Resource-Effi cient Federated Learning in Mobile Edge Computing", the authors investigate the state-of-the-art resource optimization approaches and establish a novel framework for resource efficient federated learning in mobile edge computing. In the proposed module-based federated learning framework with neural-structure-aware resource management, mobile clients are assigned with different subnetworks of the global model according to the status of their local resources. The proposed approach off ers an elastic learning framework for emerging applications such as edge computing and Internet of Things, where the clients have limited computation, bandwidth, power, and data resources. The information-theoretical perspective and module-aware approach in this work are inspiring thoughts for the research area of federated learning and edge services. In the third article,
doi:10.1109/mnet.2021.9355037 fatcat:xu2znnsebnbnpmwhjh7qcm6rgq